Recording quality

Good audio setup

The quality of every transcription, speaker separation, and text analysis depends directly on the audio quality of the recording. That’s why you should use good, unobtrusive, portable microphones for every session.

Guiding principle: High-quality clinical AI requires high-quality clinical audio.

Why this is crucial

The Quality Chain

Each stage of this pipeline amplifies or accumulates errors from the previous stage. Poor input audio quality degrades all subsequent outputs— regardless of the quality of the models used.

🎙️ Audio
Transcription
Speaker separation
Text Analysis
Documentation
AI Assistant

Poor audio quality at the input stage → errors in transcription → incorrect speaker assignments → unreliable documentation → limited AI quality.

The two main reasons

Why Wearable Microphones

The requirements in therapeutic settings differ fundamentally from those in meetings or call centers.

🤝

Therapeutic Naturalness

Technology must not dominate the therapeutic relationship

Large, conspicuous recording devices on the table can affect the atmosphere of the session. Many clients feel as though they are being watched or “recorded” which undermines the natural openness of the conversation.

Small, body-worn microphones become unnoticeable after just a few minutes. The meeting remains the focus not the technology.

Wearable microphone

  • Forgotten after 2–5 minutes
  • No technical equipment in the room
  • Natural conversation
  • The client does not feel monitored

Table or room solution

  • Always visible and present
  • Indicates "Recording in progress"
  • Changes the dynamics of the conversation
  • Reduces openness
🔊

Reliable voice separation

Software-based diarization alone is not sufficient for clinical practice

Modern diarization models are powerful but they have a crucial limitation: They can only separate what is acoustically separable. In typical therapy rooms without single-track recordings, software-only diarization fails to:

  • Overlaps and interruptions
  • Significant differences in volume (therapist speaking loudly, client speaking quietly)
  • Room reverb that blends the two voices
  • Emotional moments (crying, speaking softly, pauses)

Separate wireless microphones solve this problem physically not algorithmically. Each speaker has their own signal. The result: precise speaker labels, accurate attributions in the transcript, and reliable documentation.

Recommended devices

Wireless Wearable Setups

For clinical use, we recommend the following wireless systems one transmitter per person, worn on the body (clip-on/lavalier).

Device Why is it recommended?
RØDE Wireless GO II Top Choice Excellent voice quality, dual-channel (two transmitters simultaneously), very discreet, proven in therapeutic use. USB-C, easy setup.
DJI Mic 2 Very small magnetic transmitter, easy pairing, solid voice quality unobtrusive in everyday clinical practice.
Hollyland Lark M2 Extremely discreet and lightweight ideal when discretion is a top priority. Includes two transmitters.
RØDE Wireless ME Extremely easy to use, compact, and capable of direct connection without a separate receiver making it hassle-free for everyday use in clinical settings.

All of the devices listed are recognized as separate audio sources and can be selected directly in the app's Recording tab.

Unsupported setups

What doesn't work

The following configurations do not produce clinically useful transcriptions and are not supported by us.

Setup Best Practices

Before the meeting

Six simple steps for consistent recording quality.

Privacy & Security

What happens to the audio data

Audio data is the most sensitive data in the system and we treat it accordingly.

🔒 Audio Privacy

  • Audio files are deleted immediately after transcription no raw audio remains on our servers
  • Transcription takes place locally in the browser audio data never leaves your device in unencrypted form
  • Only the encrypted text transcript is stored end-to-end encrypted and readable only by you
  • No hidden background recording recording is only active when you explicitly start it
  • The client's consent is required prior to each recording
Complete privacy policy documentation →